intridea/grape

9 years ago

admin

42 minutes

What is Grape?

Grape is a REST-like API micro-framework for Ruby. It’s designed to run on Rack or complement existing web application frameworks such as Rails and Sinatra by providing a simple DSL to easily develop RESTful APIs. It has built-in support for common conventions, including multiple formats, subdomain/prefix restriction, content negotiation, versioning and much more.

Stable Release

You’re reading the documentation for the next release of Grape, which should be 0.11.1. Please read UPGRADING when upgrading from a previous version. The current stable release is 0.11.0.

Project Resources

Need help? Grape Google Group
Grape Wiki

Installation

Grape is available as a gem, to install it just install the gem:

gem install grape

If you’re using Bundler, add the gem to Gemfile.

gem 'grape'

Run bundle install.

Basic Usage

Grape APIs are Rack applications that are created by subclassing Grape::API. Below is a simple example showing some of the more common features of Grape in the context of recreating parts of the Twitter API.

module Twitter
  class API < Grape::API
    version 'v1', using: :header, vendor: 'twitter'
    format :json
    prefix :api

    helpers do
      def current_user
        @current_user ||= User.authorize!(env)
      end

      def authenticate!
        error!('401 Unauthorized', 401) unless current_user
      end
    end

    resource :statuses do
      desc "Return a public timeline."
      get :public_timeline do
        Status.limit(20)
      end

      desc "Return a personal timeline."
      get :home_timeline do
        authenticate!
        current_user.statuses.limit(20)
      end

      desc "Return a status."
      params do
        requires :id, type: Integer, desc: "Status id."
      end
      route_param :id do
        get do
          Status.find(params[:id])
        end
      end

      desc "Create a status."
      params do
        requires :status, type: String, desc: "Your status."
      end
      post do
        authenticate!
        Status.create!({
          user: current_user,
          text: params[:status]
        })
      end

      desc "Update a status."
      params do
        requires :id, type: String, desc: "Status ID."
        requires :status, type: String, desc: "Your status."
      end
      put ':id' do
        authenticate!
        current_user.statuses.find(params[:id]).update({
          user: current_user,
          text: params[:status]
        })
      end

      desc "Delete a status."
      params do
        requires :id, type: String, desc: "Status ID."
      end
      delete ':id' do
        authenticate!
        current_user.statuses.find(params[:id]).destroy
      end
    end
  end
end

Mounting

Rack

The above sample creates a Rack application that can be run from a rackup config.ru file with rackup:

run Twitter::API

And would respond to the following routes:

GET /api/statuses/public_timeline
GET /api/statuses/home_timeline
GET /api/statuses/:id
POST /api/statuses
PUT /api/statuses/:id
DELETE /api/statuses/:id

Grape will also automatically respond to HEAD and OPTIONS for all GET, and just OPTIONS for all other routes.

ActiveRecord without Rails

If you want to use ActiveRecord within Grape, you will need to make sure that ActiveRecord’s connection pool is handled correctly.

The easiest way to achieve that is by using ActiveRecord’s ConnectionManagement middleware in your config.ru before mounting Grape, e.g.:

use ActiveRecord::ConnectionAdapters::ConnectionManagement

run Twitter::API

Alongside Sinatra (or other frameworks)

If you wish to mount Grape alongside another Rack framework such as Sinatra, you can do so easily using Rack::Cascade:

# Example config.ru

require 'sinatra'
require 'grape'

class API < Grape::API
  get :hello do
    { hello: "world" }
  end
end

class Web < Sinatra::Base
  get '/' do
    "Hello world."
  end
end

use Rack::Session::Cookie
run Rack::Cascade.new [API, Web]

Rails

Place API files into app/api. Rails expects a subdirectory that matches the name of the Ruby module and a file name that matches the name of the class. In our example, the file name location and directory for Twitter::API should be app/api/twitter/api.rb.

Modify application.rb:

config.paths.add File.join('app', 'api'), glob: File.join('**', '*.rb')
config.autoload_paths += Dir[Rails.root.join('app', 'api', '*')]

Modify config/routes:

mount Twitter::API => '/'

Additionally, if the version of your Rails is 4.0+ and the application uses the default model layer of ActiveRecord, you will want to use the hashie-forbidden_attributes gem. This gem disables the security feature of strong_params at the model layer, allowing you the use of Grape’s own params validation instead.

# Gemfile
gem "hashie-forbidden_attributes"

See below for additional code that enables reloading of API changes in development.

Modules

You can mount multiple API implementations inside another one. These don’t have to be different versions, but may be components of the same API.

class Twitter::API < Grape::API
  mount Twitter::APIv1
  mount Twitter::APIv2
end

You can also mount on a path, which is similar to using prefix inside the mounted API itself.

class Twitter::API < Grape::API
  mount Twitter::APIv1 => '/v1'
end

Versioning

There are four strategies in which clients can reach your API’s endpoints: :path, :header, :accept_version_header and :param. The default strategy is :path.

Path

version 'v1', using: :path

Using this versioning strategy, clients should pass the desired version in the URL.

curl http://localhost:9292/v1/statuses/public_timeline

Header

version 'v1', using: :header, vendor: 'twitter'

Using this versioning strategy, clients should pass the desired version in the HTTP Accept head.

curl -H Accept:application/vnd.twitter-v1+json http://localhost:9292/statuses/public_timeline

By default, the first matching version is used when no Accept header is supplied. This behavior is similar to routing in Rails. To circumvent this default behavior, one could use the :strict option. When this option is set to true, a 406 Not Acceptable error is returned when no correct Accept header is supplied.

When an invalid Accept header is supplied, a 406 Not Acceptable error is returned if the :cascade option is set to false. Otherwise a 404 Not Found error is returned by Rack if no other route matches.

HTTP Status Code

By default Grape returns a 200 status code for GET-Requests and 201 for POST-Requests. You can use status to query and set the actual HTTP Status Code

post do
  status 202

  if status == 200
     # do some thing
  end
end

You can also use one of status codes symbols that are provided by Rack utils

post do
  status :no_content
end

Accept-Version Header

version 'v1', using: :accept_version_header

Using this versioning strategy, clients should pass the desired version in the HTTP Accept-Version header.

curl -H "Accept-Version:v1" http://localhost:9292/statuses/public_timeline

By default, the first matching version is used when no Accept-Version header is supplied. This behavior is similar to routing in Rails. To circumvent this default behavior, one could use the :strict option. When this option is set to true, a 406 Not Acceptable error is returned when no correct Accept header is supplied.

Param

version 'v1', using: :param

Using this versioning strategy, clients should pass the desired version as a request parameter, either in the URL query string or in the request body.

curl http://localhost:9292/statuses/public_timeline?apiver=v1

The default name for the query parameter is ‘apiver’ but can be specified using the :parameter option.

version 'v1', using: :param, parameter: "v"

curl http://localhost:9292/statuses/public_timeline?v=v1

Describing Methods

You can add a description to API methods and namespaces.

desc "Returns your public timeline." do
  detail 'more details'
  params  API::Entities::Status.documentation
  success API::Entities::Entity
  failure [[401, 'Unauthorized', "Entities::Error"]]
  named 'My named route'
  headers [XAuthToken: {
             description: 'Valdates your identity',
             required: true
           },
           XOptionalHeader: {
             description: 'Not really needed',
            required: false
           }
          ]
end
get :public_timeline do
  Status.limit(20)
end

detail: A more enhanced description
params: Define parameters directly from an Entity
success: (former entity) The Entity to be used to present by default this route
failure: (former http_codes) A definition of the used failure HTTP Codes and Entities
named: A helper to give a route a name and find it with this name in the documentation Hash
headers: A definition of the used Headers

Parameters

Request parameters are available through the params hash object. This includes GET, POST and PUT parameters, along with any named parameters you specify in your route strings.

get :public_timeline do
  Status.order(params[:sort_by])
end

Parameters are automatically populated from the request body on POST and PUT for form input, JSON and XML content-types.

The request:

curl -d '{"text": "140 characters"}' 'http://localhost:9292/statuses' -H Content-Type:application/json -v

The Grape endpoint:

post '/statuses' do
  Status.create!(text: params[:text])
end

Multipart POSTs and PUTs are supported as well.

The request:

curl --form image_file='@image.jpg;type=image/jpg' http://localhost:9292/upload

The Grape endpoint:

post "upload" do
  # file in params[:image_file]
end

In the case of conflict between either of:

route string parameters
GET, POST and PUT parameters
the contents of the request body on POST and PUT

route string parameters will have precedence.

Declared

Grape allows you to access only the parameters that have been declared by your params block. It filters out the params that have been passed, but are not allowed. Let’s have the following api:

format :json

post 'users/signup' do
  { "declared_params" => declared(params) }
end

If we do not specify any params, declared will return an empty Hashie::Mash instance.

Request

curl -X POST -H "Content-Type: application/json" localhost:9292/users/signup -d '{"user": {"first_name":"first name", "last_name": "last name"}}'

Response

{
  "declared_params": {}
}

Once we add parameters requirements, grape will start returning only the declared params.

format :json

params do
  requires :user, type: Hash do
    requires :first_name, type: String
    requires :last_name, type: String
  end
end

post 'users/signup' do
  { "declared_params" => declared(params) }
end

Request

curl -X POST -H "Content-Type: application/json" localhost:9292/users/signup -d '{"user": {"first_name":"first name", "last_name": "last name", "random": "never shown"}}'

Response

{
  "declared_params": {
    "user": {
      "first_name": "first name",
      "last_name": "last name"
    }
  }
}

Returned hash is a Hashie::Mash instance so you can access parameters via dot notation:

  declared(params).user == declared(params)["user"]

Include missing

By default declared(params) returns parameters that has nil value. If you want to return only the parameters that have any value, you can use the include_missing option. By default it is true. Let’s have the following api:

format :json

params do
  requires :first_name, type: String
  optional :last_name, type: String
end

post 'users/signup' do
  { "declared_params" => declared(params, include_missing: false) }
end

Request

curl -X POST -H "Content-Type: application/json" localhost:9292/users/signup -d '{"user": {"first_name":"first name", "random": "never shown"}}'

Response with include_missing:false

{
  "declared_params": {
    "user": {
      "first_name": "first name"
    }
  }
}

Response with include_missing:true

{
  "declared_params": {
    "first_name": "first name",
    "last_name": null
  }
}

It also works on nested hashes:

format :json

params do
  requires :user, :type => Hash do
    requires :first_name, type: String
    optional :last_name, type: String
    requires :address, :type => Hash do
      requires :city, type: String
      optional :region, type: String
    end
  end
end

post 'users/signup' do
  { "declared_params" => declared(params, include_missing: false) }
end

Request

curl -X POST -H "Content-Type: application/json" localhost:9292/users/signup -d '{"user": {"first_name":"first name", "random": "never shown", "address": { "city": "SF"}}}'

Response with include_missing:false

{
  "declared_params": {
    "user": {
      "first_name": "first name",
      "address": {
        "city": "SF"
      }
    }
  }
}

Response with include_missing:true

{
  "declared_params": {
    "user": {
      "first_name": "first name",
      "last_name": null,
      "address": {
        "city": "Zurich",
        "region": null
      }
    }
  }
}

Note that an attribute with a nil value is not considered missing and will also be returned when include_missing is set to false:

Request

curl -X POST -H "Content-Type: application/json" localhost:9292/users/signup -d '{"user": {"first_name":"first name", "last_name": null, "address": { "city": "SF"}}}'

Response with include_missing:false

{
  "declared_params": {
    "user": {
      "first_name": "first name",
      "last_name": null,
      "address": { "city": "SF"}
    }
  }
}

Parameter Validation and Coercion

You can define validations and coercion options for your parameters using a params block.

params do
  requires :id, type: Integer
  optional :text, type: String, regexp: /^[a-z]+$/
  group :media do
    requires :url
  end
  optional :audio do
    requires :format, type: Symbol, values: [:mp3, :wav, :aac, :ogg], default: :mp3
  end
  mutually_exclusive :media, :audio
end
put ':id' do
  # params[:id] is an Integer
end

When a type is specified an implicit validation is done after the coercion to ensure the output type is the one declared.

Optional parameters can have a default value.

params do
  optional :color, type: String, default: 'blue'
  optional :random_number, type: Integer, default: -> { Random.rand(1..100) }
  optional :non_random_number, type: Integer, default:  Random.rand(1..100)
end

Note that default values will be passed through to any validation options specified. The following example will always fail if :color is not explicitly provided.

params do
  optional :color, type: String, default: 'blue', values: ['red', 'green']
end

The correct implementation is to ensure the default value passes all validations.

params do
  optional :color, type: String, default: 'blue', values: ['blue', 'red', 'green']
end

Validation of Nested Parameters

Parameters can be nested using group or by calling requires or optional with a block. In the above example, this means params[:media][:url] is required along with params[:id], and params[:audio][:format] is required only if params[:audio] is present. With a block, group, requires and optional accept an additional option type which can be either Array or Hash, and defaults to Array. Depending on the value, the nested parameters will be treated either as values of a hash or as values of hashes in an array.

params do
  optional :preferences, type: Array do
    requires :key
    requires :value
  end

  requires :name, type: Hash do
    requires :first_name
    requires :last_name
  end
end

Built-in Validators

`allow_blank`

Parameters can be defined as allow_blank, ensuring that they contain a value. By default, requires only validates that a parameter was sent in the request, regardless its value. With allow_blank: false, empty values or whitespace only values are invalid.

allow_blank can be combined with both requires and optional. If the parameter is required, it has to contain a value. If it’s optional, it’s possible to not send it in the request, but if it’s being sent, it has to have some value, and not an empty string/only whitespaces.

params do
  requires :username, allow_blank: false
  optional :first_name, allow_blank: false
end

`values`

Parameters can be restricted to a specific set of values with the :values option.

Default values are eagerly evaluated. Above :non_random_number will evaluate to the same number for each call to the endpoint of this params block. To have the default evaluate lazily with each request use a lambda, like :random_number above.

params do
  requires :status, type: Symbol, values: [:not_started, :processing, :done]
  optional :numbers, type: Array[Integer], default: 1, values: [1, 2, 3, 5, 8]
end

Supplying a range to the :values option ensures that the parameter is (or parameters are) included in that range (using Range#include?).

params do
  requires :latitude, type: Float, values: -90.0..+90.0
  requires :longitude, type: Float, values: -180.0..+180.0
  optional :letters, type: Array[String], values: 'a'..'z'
end

Note that both range endpoints have to be a #kind_of? your :type option (if you don’t supplied the :type option, it will be guessed to be equal to the class of the range’s first endpoint). So the following is invalid:

params do
  requires :invalid1, type: Float, values: 0..10 # 0.kind_of?(Float) => false
  optional :invalid2, values: 0..10.0 # 10.0.kind_of?(0.class) => false
end

The :values option can also be supplied with a Proc, evaluated lazily with each request. For example, given a status model you may want to restrict by hashtags that you have previously defined in the HashTag model.

params do
  requires :hashtag, type: String, values: -> { Hashtag.all.map(&:tag) }
end

`regexp`

Parameters can be restricted to match a specific regular expression with the :regexp option. If the value does not match the regular expression an error will be returned. Note that this is true for both requires and optional parameters.

params do
  requires :email, regexp: /.+@.+/
end

The validator will pass if the parameter was sent without value. To ensure that the parameter contains a value, use allow_blank: false.

params do
  requires :email, allow_blank: false, regexp: /.+@.+/
end

`mutually_exclusive`

Parameters can be defined as mutually_exclusive, ensuring that they aren’t present at the same time in a request.

params do
  optional :beer
  optional :wine
  mutually_exclusive :beer, :wine
end

Multiple sets can be defined:

params do
  optional :beer
  optional :wine
  mutually_exclusive :beer, :wine
  optional :scotch
  optional :aquavit
  mutually_exclusive :scotch, :aquavit
end

Warning: Never define mutually exclusive sets with any required params. Two mutually exclusive required params will mean params are never valid, thus making the endpoint useless. One required param mutually exclusive with an optional param will mean the latter is never valid.

`exactly_one_of`

Parameters can be defined as ‘exactly_one_of’, ensuring that exactly one parameter gets selected.

params do
  optional :beer
  optional :wine
  exactly_one_of :beer, :wine
end

`at_least_one_of`

Parameters can be defined as ‘at_least_one_of’, ensuring that at least one parameter gets selected.

params do
  optional :beer
  optional :wine
  optional :juice
  at_least_one_of :beer, :wine, :juice
end

`all_or_none_of`

Parameters can be defined as ‘all_or_none_of’, ensuring that all or none of parameters gets selected.

params do
  optional :beer
  optional :wine
  optional :juice
  all_or_none_of :beer, :wine, :juice
end

Nested `mutually_exclusive`, `exactly_one_of`, `at_least_one_of`, `all_or_none_of`

All of these methods can be used at any nested level.

params do
  requires :food do
    optional :meat
    optional :fish
    optional :rice
    at_least_one_of :meat, :fish, :rice
  end
  group :drink do
    optional :beer
    optional :wine
    optional :juice
    exactly_one_of :beer, :wine, :juice
  end
  optional :dessert do
    optional :cake
    optional :icecream
    mutually_exclusive :cake, :icecream
  end
  optional :recipe do
    optional :oil
    optional :meat
    all_or_none_of :oil, :meat
  end
end

Namespace Validation and Coercion

Namespaces allow parameter definitions and apply to every method within the namespace.

namespace :statuses do
  params do
    requires :user_id, type: Integer, desc: "A user ID."
  end
  namespace ":user_id" do
    desc "Retrieve a user's status."
    params do
      requires :status_id, type: Integer, desc: "A status ID."
    end
    get ":status_id" do
      User.find(params[:user_id]).statuses.find(params[:status_id])
    end
  end
end

The namespace method has a number of aliases, including: group, resource, resources, and segment. Use whichever reads the best for your API.

You can conveniently define a route parameter as a namespace using route_param.

namespace :statuses do
  route_param :id do
    desc "Returns all replies for a status."
    get 'replies' do
      Status.find(params[:id]).replies
    end
    desc "Returns a status."
    get do
      Status.find(params[:id])
    end
  end
end

Custom Validators

class AlphaNumeric < Grape::Validations::Base
  def validate_param!(attr_name, params)
    unless params[attr_name] =~ /^[[:alnum:]]+$/
      fail Grape::Exceptions::Validation, params: [@scope.full_name(attr_name)], message: "must consist of alpha-numeric characters"
    end
  end
end

params do
  requires :text, alpha_numeric: true
end

You can also create custom classes that take parameters.

class Length < Grape::Validations::Base
  def validate_param!(attr_name, params)
    unless params[attr_name].length  'Invalid token.'

Routes

Optionally, you can define requirements for your named route parameters using regular expressions on namespace or endpoint. The route will match only if all requirements are met.

get ':id', requirements: { id: /[0-9]*/ } do
  Status.find(params[:id])
end

namespace :outer, requirements: { id: /[0-9]*/ } do
  get :id do
  end

  get ":id/edit" do
  end
end

Helpers

You can define helper methods that your endpoints can use with the helpers macro by either giving a block or a module.

module StatusHelpers
  def user_info(user)
    "#{user} has statused #{user.statuses} status(s)"
  end
end

class API < Grape::API
  # define helpers with a block
  helpers do
    def current_user
      User.find(params[:user_id])
    end
  end

  # or mix in a module
  helpers StatusHelpers

  get 'info' do
    # helpers available in your endpoint and filters
    user_info(current_user)
  end
end

You can define reusable params using helpers.

class API < Grape::API
  helpers do
    params :pagination do
      optional :page, type: Integer
      optional :per_page, type: Integer
    end
  end

  desc "Get collection"
  params do
    use :pagination # aliases: includes, use_scope
  end
  get do
    Collection.page(params[:page]).per(params[:per_page])
  end
end

You can also define reusable params using shared helpers.

module SharedParams
  extend Grape::API::Helpers

  params :period do
    optional :start_date
    optional :end_date
  end

  params :pagination do
    optional :page, type: Integer
    optional :per_page, type: Integer
  end
end

class API < Grape::API
  helpers SharedParams

  desc "Get collection."
  params do
    use :period, :pagination
  end

  get do
    Collection
      .from(params[:start_date])
      .to(params[:end_date])
      .page(params[:page])
      .per(params[:per_page])
  end
end

Helpers support blocks that can help set default values. The following API can return a collection sorted by id or created_at in asc or desc order.

module SharedParams
  extend Grape::API::Helpers

  params :order do |options|
    optional :order_by, type:Symbol, values:options[:order_by], default:options[:default_order_by]
    optional :order, type:Symbol, values:%i(asc desc), default:options[:default_order]
  end
end

class API < Grape::API
  helpers SharedParams

  desc "Get a sorted collection."
  params do
    use :order, order_by:%i(id created_at), default_order_by: :created_at, default_order: :asc
  end

  get do
    Collection.send(params[:order], params[:order_by])
  end
end

Parameter Documentation

You can attach additional documentation to params using a documentation hash.

params do
  optional :first_name, type: String, documentation: { example: 'Jim' }
  requires :last_name, type: String, documentation: { example: 'Smith' }
end

Cookies

You can set, get and delete your cookies very simply using cookies method.

class API < Grape::API
  get 'status_count' do
    cookies[:status_count] ||= 0
    cookies[:status_count] += 1
    { status_count: cookies[:status_count] }
  end

  delete 'status_count' do
    { status_count: cookies.delete(:status_count) }
  end
end

Use a hash-based syntax to set more than one value.

cookies[:status_count] = {
  value: 0,
  expires: Time.tomorrow,
  domain: '.twitter.com',
  path: '/'
}

cookies[:status_count][:value] +=1

Delete a cookie with delete.

cookies.delete :status_count

Specify an optional path.

cookies.delete :status_count, path: '/'

Redirecting

You can redirect to a new url temporarily (302) or permanently (301).

redirect '/statuses'

redirect '/statuses', permanent: true

Allowed Methods

When you add a GET route for a resource, a route for the HEAD method will also be added automatically. You can disable this behavior with do_not_route_head!.

class API < Grape::API
  do_not_route_head!

  get '/example' do
    # only responds to GET
  end
end

When you add a route for a resource, a route for the OPTIONS method will also be added. The response to an OPTIONS request will include an “Allow” header listing the supported methods.

class API < Grape::API
  get '/rt_count' do
    { rt_count: current_user.rt_count }
  end

  params do
    requires :value, type: Integer, desc: 'Value to add to the rt count.'
  end
  put '/rt_count' do
    current_user.rt_count += params[:value].to_i
    { rt_count: current_user.rt_count }
  end
end

curl -v -X OPTIONS http://localhost:3000/rt_count

> OPTIONS /rt_count HTTP/1.1
>
< HTTP/1.1 204 No Content
< Allow: OPTIONS, GET, PUT

You can disable this behavior with do_not_route_options!.

If a request for a resource is made with an unsupported HTTP method, an HTTP 405 (Method Not Allowed) response will be returned.

curl -X DELETE -v http://localhost:3000/rt_count/

> DELETE /rt_count/ HTTP/1.1
> Host: localhost:3000
>
< HTTP/1.1 405 Method Not Allowed
< Allow: OPTIONS, GET, PUT

Raising Exceptions

You can abort the execution of an API method by raising errors with error!.

error! 'Access Denied', 401

You can also return JSON formatted objects by raising error! and passing a hash instead of a message.

error!({ error: "unexpected error", detail: "missing widget" }, 500)

You can present documented errors with a Grape entity using the the grape-entity gem.

module API
  class Error < Grape::Entity
    expose :code
    expose :message
  end
end

The following example specifies the entity to use in the http_codes definition.

desc 'My Route' do
 failure [[408, 'Unauthorized', API::Error]]
end
error!({ message: 'Unauthorized' }, 408)

The following example specifies the presented entity explicitly in the error message.

desc 'My Route' do
 failure [[408, 'Unauthorized']]
end
error!({ message: 'Unauthorized', with: API::Error }, 408)

Default Error HTTP Status Code

By default Grape returns a 500 status code from error!. You can change this with default_error_status.

class API < Grape::API
  default_error_status 400
  get '/example' do
    error! "This should have http status code 400"
  end
end

Handling 404

For Grape to handle all the 404s for your API, it can be useful to use a catch-all. In its simplest form, it can be like:

route :any, '*path' do
  error! # or something else
end

It is very crucial to define this endpoint at the very end of your API, as it literally accepts every request.

Exception Handling

Grape can be told to rescue all exceptions and return them in the API format.

class Twitter::API < Grape::API
  rescue_from :all
end

You can also rescue specific exceptions.

class Twitter::API < Grape::API
  rescue_from ArgumentError, UserDefinedError
end

In this case UserDefinedError must be inherited from StandardError.

The error format will match the request format. See “Content-Types” below.

Custom error formatters for existing and additional types can be defined with a proc.

class Twitter::API < Grape::API
  error_formatter :txt, lambda { |message, backtrace, options, env|
    "error: #{message} from #{backtrace}"
  }
end

You can also use a module or class.

module CustomFormatter
  def self.call(message, backtrace, options, env)
    { message: message, backtrace: backtrace }
  end
end

class Twitter::API < Grape::API
  error_formatter :custom, CustomFormatter
end

You can rescue all exceptions with a code block. The error! wrapper automatically sets the default error code and content-type.

class Twitter::API < Grape::API
  rescue_from :all do |e|
    error!("rescued from #{e.class.name}")
  end
end

Optionally, you can set the format, status code and headers.

class Twitter::API < Grape::API
  format :json
  rescue_from :all do |e|
    error!({ error: "Server error.", 500, { 'Content-Type' => 'text/error' } })
  end
end

You can also rescue specific exceptions with a code block and handle the Rack response at the lowest level.

class Twitter::API < Grape::API
  rescue_from :all do |e|
    Rack::Response.new([ e.message ], 500, { "Content-type" => "text/error" }).finish
  end
end

Or rescue specific exceptions.

class Twitter::API < Grape::API
  rescue_from ArgumentError do |e|
    error!("ArgumentError: #{e.message}")
  end

  rescue_from NotImplementedError do |e|
    error!("NotImplementedError: #{e.message}")
  end
end

By default, rescue_from will rescue the exceptions listed and all their subclasses.

Assume you have the following exception classes defined.

module APIErrors
  class ParentError < StandardError; end
  class ChildError < ParentError; end
end

Then the following rescue_from clause will rescue exceptions of type APIErrors::ParentError and its subclasses (in this case APIErrors::ChildError).

rescue_from APIErrors::ParentError do |e|
    error!({
      error: "#{e.class} error",
      message: e.message
    }, e.status)
end

To only rescue the base exception class, set rescue_subclasses: false. The code below will rescue exceptions of type RuntimeError but not its subclasses.

rescue_from RuntimeError, rescue_subclasses: false do |e|
    error!({
      status: e.status,
      message: e.message,
      errors: e.errors
    }, e.status)
end

Rails 3.x

When mounted inside containers, such as Rails 3.x, errors like “404 Not Found” or “406 Not Acceptable” will likely be handled and rendered by Rails handlers. For instance, accessing a nonexistent route “/api/foo” raises a 404, which inside rails will ultimately be translated to an ActionController::RoutingError, which most likely will get rendered to a HTML error page.

Most APIs will enjoy preventing downstream handlers from handling errors. You may set the :cascade option to false for the entire API or separately on specific version definitions, which will remove the X-Cascade: true header from API responses.

cascade false

version 'v1', using: :header, vendor: 'twitter', cascade: false

Logging

Grape::API provides a logger method which by default will return an instance of the Logger class from Ruby’s standard library.

To log messages from within an endpoint, you need to define a helper to make the logger available in the endpoint context.

class API < Grape::API
  helpers do
    def logger
      API.logger
    end
  end
  post '/statuses' do
    # ...
    logger.info "#{current_user} has statused"
  end
end

You can also set your own logger.

class MyLogger
  def warning(message)
    puts "this is a warning: #{message}"
  end
end

class API < Grape::API
  logger MyLogger.new
  helpers do
    def logger
      API.logger
    end
  end
  get '/statuses' do
    logger.warning "#{current_user} has statused"
  end
end

For similar to Rails request logging try the grape_logging gem.

API Formats

Your API can declare which content-types to support by using content_type. If you do not specify any, Grape will support XML, JSON, BINARY, and TXT content-types. The default format is :txt; you can change this with default_format. Essentially, the two APIs below are equivalent.

class Twitter::API < Grape::API
  # no content_type declarations, so Grape uses the defaults
end

class Twitter::API < Grape::API
  # the following declarations are equivalent to the defaults

  content_type :xml, 'application/xml'
  content_type :json, 'application/json'
  content_type :binary, 'application/octet-stream'
  content_type :txt, 'text/plain'

  default_format :txt
end

If you declare any content_type whatsoever, the Grape defaults will be overridden. For example, the following API will only support the :xml and :rss content-types, but not :txt, :json, or :binary. Importantly, this means the :txt default format is not supported! So, make sure to set a new default_format.

class Twitter::API < Grape::API
  content_type :xml, 'application/xml'
  content_type :rss, 'application/xml+rss'

  default_format :xml
end

Serialization takes place automatically. For example, you do not have to call to_json in each JSON API endpoint implementation. The response format (and thus the automatic serialization) is determined in the following order:

Use the file extension, if specified. If the file is .json, choose the JSON format.
Use the value of the format parameter in the query string, if specified.
Use the format set by the format option, if specified.
Attempt to find an acceptable format from the Accept header.
Use the default format, if specified by the default_format option.
Default to :txt.

For example, consider the following API.

class MultipleFormatAPI < Grape::API
  content_type :xml, 'application/xml'
  content_type :json, 'application/json'

  default_format :json

  get :hello do
    { hello: 'world' }
  end
end

GET /hello (with an Accept: */* header) does not have an extension or a format parameter, so it will respond with JSON (the default format).
GET /hello.xml has a recognized extension, so it will respond with XML.
GET /hello?format=xml has a recognized format parameter, so it will respond with XML.
GET /hello.xml?format=json has a recognized extension (which takes precedence over the format parameter), so it will respond with XML.
GET /hello.xls (with an Accept: */* header) has an extension, but that extension is not recognized, so it will respond with JSON (the default format).
GET /hello.xls with an Accept: application/xml header has an unrecognized extension, but the Accept header corresponds to a recognized format, so it will respond with XML.
GET /hello.xls with an Accept: text/plain header has an unrecognized extension and an unrecognized Accept header, so it will respond with JSON (the default format).

You can override this process explicitly by specifying env['api.format'] in the API itself. For example, the following API will let you upload arbitrary files and return their contents as an attachment with the correct MIME type.

class Twitter::API < Grape::API
  post "attachment" do
    filename = params[:file][:filename]
    content_type MIME::Types.type_for(filename)[0].to_s
    env['api.format'] = :binary # there's no formatter for :binary, data will be returned "as is"
    header "Content-Disposition", "attachment; filename*=UTF-8''#{URI.escape(filename)}"
    params[:file][:tempfile].read
  end
end

You can have your API only respond to a single format with format. If you use this, the API will not respond to file extensions other than specified in format. For example, consider the following API.

class SingleFormatAPI < Grape::API
  format :json

  get :hello do
    { hello: 'world' }
  end
end

GET /hello will respond with JSON.
GET /hello.json will respond with JSON.
GET /hello.xml, GET /hello.foobar, or any other extension will respond with an HTTP 404 error code.
GET /hello?format=xml will respond with an HTTP 406 error code, because the XML format specified by the request parameter is not supported.
GET /hello with an Accept: application/xml header will still respond with JSON, since it could not negotiate a recognized content-type from the headers and JSON is the effective default.

The formats apply to parsing, too. The following API will only respond to the JSON content-type and will not parse any other input than application/json, application/x-www-form-urlencoded, multipart/form-data, multipart/related and multipart/mixed. All other requests will fail with an HTTP 406 error code.

class Twitter::API < Grape::API
  format :json
end

When the content-type is omitted, Grape will return a 406 error code unless default_format is specified. The following API will try to parse any data without a content-type using a JSON parser.

class Twitter::API < Grape::API
  format :json
  default_format :json
end

If you combine format with rescue_from :all, errors will be rendered using the same format. If you do not want this behavior, set the default error formatter with default_error_formatter.

class Twitter::API < Grape::API
  format :json
  content_type :txt, "text/plain"
  default_error_formatter :txt
end

Custom formatters for existing and additional types can be defined with a proc.

class Twitter::API < Grape::API
  content_type :xls, "application/vnd.ms-excel"
  formatter :xls, lambda { |object, env| object.to_xls }
end

You can also use a module or class.

module XlsFormatter
  def self.call(object, env)
    object.to_xls
  end
end

class Twitter::API < Grape::API
  content_type :xls, "application/vnd.ms-excel"
  formatter :xls, XlsFormatter
end

Built-in formatters are the following.

:json: use object’s to_json when available, otherwise call MultiJson.dump
:xml: use object’s to_xml when available, usually via MultiXml, otherwise call to_s
:txt: use object’s to_txt when available, otherwise to_s
:serializable_hash: use object’s serializable_hash when available, otherwise fallback to :json
:binary: data will be returned “as is”

JSONP

Grape supports JSONP via Rack::JSONP, part of the rack-contrib gem. Add rack-contrib to your Gemfile.

require 'rack/contrib'

class API < Grape::API
  use Rack::JSONP
  format :json
  get '/' do
    'Hello World'
  end
end

CORS

Grape supports CORS via Rack::CORS, part of the rack-cors gem. Add rack-cors to your Gemfile, then use the middleware in your config.ru file.

require 'rack/cors'

use Rack::Cors do
  allow do
    origins '*'
    resource '*', headers: :any, methods: :get
  end
end

run Twitter::API

Content-type

Content-type is set by the formatter. You can override the content-type of the response at runtime by setting the Content-Type header.

class API < Grape::API
  get '/home_timeline_js' do
    content_type "application/javascript"
    "var statuses = ...;"
  end
end

API Data Formats

Grape accepts and parses input data sent with the POST and PUT methods as described in the Parameters section above. It also supports custom data formats. You must declare additional content-types via content_type and optionally supply a parser via parser unless a parser is already available within Grape to enable a custom format. Such a parser can be a function or a class.

With a parser, parsed data is available “as-is” in env['api.request.body']. Without a parser, data is available “as-is” and in env['api.request.input'].

The following example is a trivial parser that will assign any input with the “text/custom” content-type to :value. The parameter will be available via params[:value] inside the API call.

module CustomParser
  def self.call(object, env)
    { value: object.to_s }
  end
end

content_type :txt, "text/plain"
content_type :custom, "text/custom"
parser :custom, CustomParser

put "value" do
  params[:value]
end

You can invoke the above API as follows.

curl -X PUT -d 'data' 'http://localhost:9292/value' -H Content-Type:text/custom -v

You can disable parsing for a content-type with nil. For example, parser :json, nil will disable JSON parsing altogether. The request data is then available as-is in env['api.request.body'].

RESTful Model Representations

Grape supports a range of ways to present your data with some help from a generic present method, which accepts two arguments: the object to be presented and the options associated with it. The options hash may include :with, which defines the entity to expose.

Grape Entities

Add the grape-entity gem to your Gemfile. Please refer to the grape-entity documentation for more details.

The following example exposes statuses.

module API
  module Entities
    class Status < Grape::Entity
      expose :user_name
      expose :text, documentation: { type: "string", desc: "Status update text." }
      expose :ip, if: { type: :full }
      expose :user_type, :user_id, if: lambda { |status, options| status.user.public? }
      expose :digest { |status, options| Digest::MD5.hexdigest(status.txt) }
      expose :replies, using: API::Status, as: :replies
    end
  end

  class Statuses < Grape::API
    version 'v1'

    desc 'Statuses index' do
      params: API::Entities::Status.documentation
    end
    get '/statuses' do
      statuses = Status.all
      type = current_user.admin? ? :full : :default
      present statuses, with: API::Entities::Status, type: type
    end
  end
end

You can use entity documentation directly in the params block with using: Entity.documentation.

module API
  class Statuses < Grape::API
    version 'v1'

    desc 'Create a status'
    params do
      requires :all, except: [:ip], using: API::Entities::Status.documentation.except(:id)
    end
    post '/status' do
      Status.create! params
    end
  end
end

You can present with multiple entities using an optional Symbol argument.

  get '/statuses' do
    statuses = Status.all.page(1).per(20)
    present :total_page, 10
    present :per_page, 20
    present :statuses, statuses, with: API::Entities::Status
  end

The response will be

  {
    total_page: 10,
    per_page: 20,
    statuses: []
  }

In addition to separately organizing entities, it may be useful to put them as namespaced classes underneath the model they represent.

class Status
  def entity
    Entity.new(self)
  end

  class Entity < Grape::Entity
    expose :text, :user_id
  end
end

If you organize your entities this way, Grape will automatically detect the Entity class and use it to present your models. In this example, if you added present Status.new to your endpoint, Grape will automatically detect that there is a Status::Entity class and use that as the representative entity. This can still be overridden by using the :with option or an explicit represents call.

You can present hash with Grape::Presenters::Presenter to keep things consistent.

get '/users' do
  present { id: 10, name: :dgz }, with: Grape::Presenters::Presenter
end

The response will be

{
  id:   10,
  name: 'dgz'
}

It has the same result with

get '/users' do
  present :id, 10
  present :name, :dgz
end

Hypermedia and Roar

You can use Roar to render HAL or Collection+JSON with the help of grape-roar, which defines a custom JSON formatter and enables presenting entities with Grape’s present keyword.

Rabl

You can use Rabl templates with the help of the grape-rabl gem, which defines a custom Grape Rabl formatter.

Active Model Serializers

You can use Active Model Serializers serializers with the help of the grape-active_model_serializers gem, which defines a custom Grape AMS formatter.

Sending Raw or No Data

In general, use the binary format to send raw data.

class API < Grape::API
  get '/file' do
    content_type 'application/octet-stream'
    File.binread 'file.bin'
  end
end

You can also set the response body explicitly with body.

class API < Grape::API
  get '/' do
    content_type 'text/plain'
    body 'Hello World'
    # return value ignored
  end
end

Use body false to return 204 No Content without any data or content-type.

Authentication

Basic and Digest Auth

Grape has built-in Basic and Digest authentication (the given block is executed in the context of the current Endpoint). Authentication applies to the current namespace and any children, but not parents.

http_basic do |username, password|
  # verify user's password here
  { 'test' => 'password1' }[username] == password
end

http_digest({ realm: 'Test Api', opaque: 'app secret' }) do |username|
  # lookup the user's password here
  { 'user1' => 'password1' }[username]
end

Register custom middleware for authentication

Grape can use custom Middleware for authentication. How to implement these Middleware have a look at Rack::Auth::Basic or similar implementations.

For registering a Middleware you need the following options:

label – the name for your authenticator to use it later
MiddlewareClass – the MiddlewareClass to use for authentication
option_lookup_proc – A Proc with one Argument to lookup the options at runtime (return value is an Array as Paramter for the Middleware).

Example:


Grape::Middleware::Auth::Strategies.add(:my_auth, AuthMiddleware, ->(options) { [options[:realm]] } )


auth :my_auth, { realm: 'Test Api'} do |credentials|
  # lookup the user's password here
  { 'user1' => 'password1' }[username]
end

Use warden-oauth2 or rack-oauth2 for OAuth2 support.

Describing and Inspecting an API

Grape routes can be reflected at runtime. This can notably be useful for generating documentation.

Grape exposes arrays of API versions and compiled routes. Each route contains a route_prefix, route_version, route_namespace, route_method, route_path and route_params. You can add custom route settings to the route metadata with route_setting.

class TwitterAPI < Grape::API
  version 'v1'
  desc "Includes custom settings."
  route_setting :custom, key: 'value'
  get do

  end
end

Examine the routes at runtime.

TwitterAPI::versions # yields [ 'v1', 'v2' ]
TwitterAPI::routes # yields an array of Grape::Route objects
TwitterAPI::routes[0].route_version # => 'v1'
TwitterAPI::routes[0].route_description # => 'Includes custom settings.'
TwitterAPI::routes[0].route_settings[:custom] # => { key: 'value' }

Current Route and Endpoint

It’s possible to retrieve the information about the current route from within an API call with route.

class MyAPI < Grape::API
  desc "Returns a description of a parameter."
  params do
    requires :id, type: Integer, desc: "Identity."
  end
  get "params/:id" do
    route.route_params[params[:id]] # yields the parameter description
  end
end

The current endpoint responding to the request is self within the API block or env['api.endpoint'] elsewhere. The endpoint has some interesting properties, such as source which gives you access to the original code block of the API implementation. This can be particularly useful for building a logger middleware.

class ApiLogger < Grape::Middleware::Base
  def before
    file = env['api.endpoint'].source.source_location[0]
    line = env['api.endpoint'].source.source_location[1]
    logger.debug "[api] #{file}:#{line}"
  end
end

Before and After

Blocks can be executed before or after every API call, using before, after, before_validation and after_validation.

Before and after callbacks execute in the following order:

before
before_validation
validations
after_validation
the API call
after

Steps 4, 5 and 6 only happen if validation succeeds.

E.g. using before:

before do
  header "X-Robots-Tag", "noindex"
end

The block applies to every API call within and below the current namespace:

class MyAPI < Grape::API
  get '/' do
    "root - #{@blah}"
  end

  namespace :foo do
    before do
      @blah = 'blah'
    end

    get '/' do
      "root - foo - #{@blah}"
    end

    namespace :bar do
      get '/' do
        "root - foo - bar - #{@blah}"
      end
    end
  end
end

The behaviour is then:

GET /           # 'root - '
GET /foo        # 'root - foo - blah'
GET /foo/bar    # 'root - foo - bar - blah'

Params on a namespace (or whatever alias you are using) also work when using before_validation or after_validation:

class MyAPI < Grape::API
  params do
    requires :blah, type: Integer
  end
  resource ':blah' do
    after_validation do
      # if we reach this point validations will have passed
      @blah = declared(params, include_missing: false)[:blah]
    end

    get '/' do
      @blah.class
    end
  end
end

The behaviour is then:

GET /123        # 'Fixnum'
GET /foo        # 400 error - 'blah is invalid'

When a callback is defined within a version block, it’s only called for the routes defined in that block.

class Test < Grape::API
  resource :foo do
    version 'v1', :using => :path do
      before do
        @output ||= 'v1-'
      end
      get '/' do
        @output += 'hello'
      end
    end

    version 'v2', :using => :path do
      before do
        @output ||= 'v2-'
      end
      get '/' do
        @output += 'hello'
      end
    end
  end
end

The behaviour is then:

GET /foo/v1       # 'v1-hello'
GET /foo/v2       # 'v2-hello'

Anchoring

Grape by default anchors all request paths, which means that the request URL should match from start to end to match, otherwise a 404 Not Found is returned. However, this is sometimes not what you want, because it is not always known upfront what can be expected from the call. This is because Rack-mount by default anchors requests to match from the start to the end, or not at all. Rails solves this problem by using a anchor: false option in your routes. In Grape this option can be used as well when a method is defined.

For instance when your API needs to get part of an URL, for instance:

class TwitterAPI < Grape::API
  namespace :statuses do
    get '/(*:status)', anchor: false do

    end
  end
end

This will match all paths starting with ‘/statuses/’. There is one caveat though: the params[:status] parameter only holds the first part of the request url. Luckily this can be circumvented by using the described above syntax for path specification and using the PATH_INFO Rack environment variable, using env["PATH_INFO"]. This will hold everything that comes after the ‘/statuses/’ part.

Rails Middleware

Note that when you’re using Grape mounted on Rails you don’t have to use Rails middleware because it’s already included into your middleware stack. You only have to implement the helpers to access the specific env variable.

Remote IP

By default you can access remote IP with request.ip. This is the remote IP address implemented by Rack. Sometimes it is desirable to get the remote IP Rails-style with ActionDispatch::RemoteIp.

Add gem 'actionpack' to your Gemfile and require 'action_dispatch/middleware/remote_ip.rb'. Use the middleware in your API and expose a client_ip helper. See this documentation for additional options.

class API < Grape::API
  use ActionDispatch::RemoteIp

  helpers do
    def client_ip
      env["action_dispatch.remote_ip"].to_s
    end
  end

  get :remote_ip do
    { ip: client_ip }
  end
end

Writing Tests

Writing Tests with Rack

Use rack-test and define your API as app.

RSpec

You can test a Grape API with RSpec by making HTTP requests and examining the response.

require 'spec_helper'

describe Twitter::API do
  include Rack::Test::Methods

  def app
    Twitter::API
  end

  describe Twitter::API do
    describe "GET /api/statuses/public_timeline" do
      it "returns an empty array of statuses" do
        get "/api/statuses/public_timeline"
        expect(last_response.status).to eq(200)
        expect(JSON.parse(last_response.body)).to eq []
      end
    end
    describe "GET /api/statuses/:id" do
      it "returns a status by id" do
        status = Status.create!
        get "/api/statuses/#{status.id}"
        expect(last_response.body).to eq status.to_json
      end
    end
  end
end

Airborne

You can test with other RSpec-based frameworks, including Airborne, which uses rack-test to make requests.

require 'airborne'

Airborne.configure do |config|
  config.rack_app = Twitter::API
end

describe Twitter::API do
  describe "GET /api/statuses/:id" do
    it "returns a status by id" do
      status = Status.create!
      get "/api/statuses/#{status.id}"
      expect_json(status.as_json)
    end
  end
end

MiniTest

require "test_helper"

class Twitter::APITest < MiniTest::Test
  include Rack::Test::Methods

  def app
    Twitter::API
  end

  def test_get_api_statuses_public_timeline_returns_an_empty_array_of_statuses
    get "/api/statuses/public_timeline"
    assert last_response.ok?
    assert_equal [], JSON.parse(last_response.body)
  end

  def test_get_api_statuses_id_returns_a_status_by_id
    status = Status.create!
    get "/api/statuses/#{status.id}"
    assert_equal status.to_json, last_response.body
  end
end

Writing Tests with Rails

RSpec

describe Twitter::API do
  describe "GET /api/statuses/public_timeline" do
    it "returns an empty array of statuses" do
      get "/api/statuses/public_timeline"
      expect(response.status).to eq(200)
      expect(JSON.parse(response.body)).to eq []
    end
  end
  describe "GET /api/statuses/:id" do
    it "returns a status by id" do
      status = Status.create!
      get "/api/statuses/#{status.id}"
      expect(response.body).to eq status.to_json
    end
  end
end

In Rails, HTTP request tests would go into the spec/requests group. You may want your API code to go into app/api – you can match that layout under spec by adding the following in spec/spec_helper.rb.

RSpec.configure do |config|
  config.include RSpec::Rails::RequestExampleGroup, type: :request, file_path: /spec\/api/
end

MiniTest

class Twitter::APITest < ActiveSupport::TestCase
  include Rack::Test::Methods

  def app
    Rails.application
  end

  test "GET /api/statuses/public_timeline returns an empty array of statuses" do
    get "/api/statuses/public_timeline"
    assert last_response.ok?
    assert_equal [], JSON.parse(last_response.body)
  end

  test "GET /api/statuses/:id returns a status by id" do
    status = Status.create!
    get "/api/statuses/#{status.id}"
    assert_equal status.to_json, last_response.body
  end
end

Stubbing Helpers

Because helpers are mixed in based on the context when an endpoint is defined, it can be difficult to stub or mock them for testing. The Grape::Endpoint.before_each method can help by allowing you to define behavior on the endpoint that will run before every request.

describe 'an endpoint that needs helpers stubbed' do
  before do
    Grape::Endpoint.before_each do |endpoint|
      allow(endpoint).to receive(:helper_name).and_return('desired_value')
    end
  end

  after do
    Grape::Endpoint.before_each nil
  end

  it 'should properly stub the helper' do
    # ...
  end
end

Reloading API Changes in Development

Reloading in Rack Applications

Use grape-reload.

Reloading in Rails Applications

Add API paths to config/application.rb.

# Auto-load API and its subdirectories
config.paths.add File.join("app", "api"), glob: File.join("**", "*.rb")
config.autoload_paths += Dir[Rails.root.join("app", "api", "*")]

Create config/initializers/reload_api.rb.

if Rails.env.development?
  ActiveSupport::Dependencies.explicitly_unloadable_constants

Table of Contents

What is Grape?

Stable Release

Project Resources

Installation

Basic Usage

Mounting

Rack

ActiveRecord without Rails

Alongside Sinatra (or other frameworks)

Rails

Modules

Versioning

Path

Header

HTTP Status Code

Accept-Version Header

Param

Describing Methods

Parameters

Declared

Include missing

Parameter Validation and Coercion

Validation of Nested Parameters

Built-in Validators

allow_blank

values

regexp

mutually_exclusive

exactly_one_of

at_least_one_of

all_or_none_of

Nested mutually_exclusive, exactly_one_of, at_least_one_of, all_or_none_of

Namespace Validation and Coercion

Custom Validators

Routes

Helpers

Parameter Documentation

Cookies

Redirecting

Allowed Methods

Raising Exceptions

Default Error HTTP Status Code

Handling 404

Exception Handling

Rails 3.x

Logging

API Formats

JSONP

CORS

Content-type

API Data Formats

RESTful Model Representations

Grape Entities

Hypermedia and Roar

Rabl

Active Model Serializers

Sending Raw or No Data

Authentication

Basic and Digest Auth

Register custom middleware for authentication

Describing and Inspecting an API

Current Route and Endpoint

Before and After

Anchoring

Rails Middleware

Remote IP

Writing Tests

Writing Tests with Rack

RSpec

Airborne

MiniTest

Writing Tests with Rails

RSpec

MiniTest

Stubbing Helpers

Reloading API Changes in Development

Reloading in Rack Applications

Reloading in Rails Applications

`allow_blank`

`values`

`regexp`

`mutually_exclusive`

`exactly_one_of`

`at_least_one_of`

`all_or_none_of`

Nested `mutually_exclusive`, `exactly_one_of`, `at_least_one_of`, `all_or_none_of`